I think will give three examples of problems with whom I stayed over longer time: Spaced repetition learning, polyphasic sleep and quantified self.
Quantified Self is the example where I have the most to show publically. I did community work in QS. My name is in a dozen mainstream media pieces in a total of three languages. Piece means either newspaper, radio or TV I did all of them multiple times.
Spaced repetition learning would be one problem which is extremly important but has very few people who are working on it.
The Mnemosyth data lies around for years without anyone analysing it. Going through that data and doing a bit of modeling with it should be easy for anyone who's searching a bachlor thesis for computer science or otherwise seeks a project.
Another question would be: How do you calculate a good brainperformance score for a given day given Anki review data? (Anki stores all the review data internally in a SQL database)
You don't need to be a genius to contribute to any of the those two issues. Both problems are pretty straightforward if you can program and have interest in modelling.
Polyphasic sleep is a problem where I would say that I contribute to the discussion. I tried it probably 8/9 years ago and I stayed with the problem intellectually. Last year a friend of mine was trying uberman for a month and in researching the topic he came about something I wrote. When talking with him about the topic he quoted one of my online opinion on the topic to me and at first it surprised me because I haven't made that point in his physical presence.
My highest rated answer on skeptic stackexchange is also about the uberman shedule: http://skeptics.stackexchange.com/questions/999/does-polyphasic-sleep-work-does-it-have-long-term-or-short-term-side-effects/1007#1007
It's not like I contributed a breakthrough in thinking about polyphasic sleep but I did contribute to the knowledge on the topic a bit.
The Mnemosyne data lies around for years without anyone analysing it. Going through that data and doing a bit of modeling with it should be easy for anyone who's searching a bachlor thesis for computer science or otherwise seeks a project.
It's a real pain to, though, because it's so big. A month after I started, I'm still only halfway through the logs->SQL step.
What should a not-very-smart person do? Suppose you know a not-very-smart person (around or below average intelligence). S/he read about rationality, has utilitarian inclinations, and wants to make the world better. However, s/he isn't smart enough to discover new knowledge in most fields, or contribute very much to a conversation of more knowledgeable experts on a given topic. Let's assume s/he has no exceptional talents in any area.
How do you think a person like that could best use his/her time and energy? What would you tell the person to do? This person may be, compared to average LW readership, less capable of noticing the irrationality in his/her actions even if s/he wants to be rid of it, and less easily able to notice the flaws in a bad argument. S/he may never be able to deeply understand why certain arguments are correct, certain scientific facts have to be the way they are, and telling him/her to be unsure or sure about anything seems dangerous if s/he doesn't really understand why.
My practical advice might be:
1) If you want to give to charity, follow GiveWell recommendations.
2) Learn about the basic biases, and commit to resisting them in your own life.
3) Follow advice that has been tested, that correctly predicts a positive outcome. If a hypothesis is untestable (there's an unsensible dragon in your garage) or doesn't predict anything (fire comes from phlogiston in combustable substances), or is tested and demonstrably false (god will smite you if you say it doesn't exist), don't waste time and energy on it. If you want to improve, look for tested methods that have significant positive results relevant to the area of interest. Similarly, if a person regularly gives you advice that does not lead to good outcomes, stop following it, and if someone gives advice that leads to good outcomes, start paying attention even if you like that person less.
At a more general level, my thoughts are tentative, but might include basic LW tenets such as:
1) Don't be afraid of the truth, because you're already enduring it.
2) If all the experts in a field agree on something, they might be wrong, but you are extremely unlikely to be better at uncovering the truth, so follow their advice, which might appear to conflict with...
3) Don't trust deep wisdom. Use Occam's razor, think about simple, basic reasons something might be true (this seems good for religion and moral issues, bad for scientific ideas and understanding)
4) If you find yourself flinching away from an idea, notice that, and give it extra attention.
Note: I mean this as a serious, and hopefully non-insulting question. Most people are intellectually near-average or below-average, and I have not seen extensive discussion on how to help them lead happier lives that make the world a better place.